MCP Server

Connect Trakkr to ChatGPT, Claude, Claude Code, Cursor, VS Code, Codex, and other MCP-compatible AI assistants. Query your visibility data conversationally - no code required.

What is MCP?

The Model Context Protocol is an open standard that lets AI assistants connect to external tools. Once Trakkr is connected, you ask questions in plain English and your assistant calls the right tools behind the scenes - backed by live data from your account.

Natural language

Ask in plain English. No endpoints, no parameters.

76 tools

Visibility, competitors, reports, crawler workflows.

Hosted or local

Hosted connectors for chat apps, local setup for developer tools.

Public Knowledge MCP

Trakkr also exposes a separate public, read-only MCP at https://api.trakkr.ai/public/mcp. It does not require OAuth, a connect token, or a Trakkr account. Use it when an assistant needs citeable Trakkr context about the product, research, pricing, setup docs and public datasets.

search and fetch

Company-knowledge style tools that return canonical Trakkr URLs for citations.

Public datasets

Whitelisted metadata for Bias in AI, AI Search Tools and selected research stats.

The public MCP cannot access private brand visibility, reports, prompts, citations, competitors, actions, or customer data. For those workflows, connect the authenticated MCP at https://api.trakkr.ai/mcp.

Trakkr developer interfaces, authentication, inputs, outputs, and uses
Interface	Authentication	Inputs	Outputs	Use it for
Public Knowledge MCP`api.trakkr.ai/public/mcp`	None	A search query, fetched content ID, dataset key, or client name	Public text, metadata, dataset links, and canonical citation URLs	Product research, public docs, citations, and dataset discovery
Authenticated MCP`api.trakkr.ai/mcp`	Hosted OAuth or an MCP connect token	Tool parameters such as brand ID, date window, view, and filters	Live account data or an explicit workspace action result	Private visibility, prompts, citations, reports, actions, and workflows
REST API`api.trakkr.ai`	`sk_live_` bearer key	HTTP path, query or JSON body, and brand-scoped identifiers	Documented JSON responses or exports	Scale and Enterprise integrations, dashboards, and scheduled data flows

Public MCP example

Call search first. Pass its result ID to fetch when you need the full text and canonical URL to cite. The example output is shortened, but keeps the real field names and value types.

search input

{
  "query": "What does Trakkr track?"
}

search output

{
  "query": "What does Trakkr track?",
  "results": [
    {
      "id": "trakkr-overview",
      "title": "What Trakkr does",
      "url": "https://trakkr.ai/",
      "kind": "overview",
      "summary": "Trakkr is an AI visibility platform for brands and agencies.",
      "tags": ["overview", "ai visibility"],
      "updated_at": "YYYY-MM-DD",
      "score": 1
    }
  ],
  "result_count": 1,
  "private_data_intent": false,
  "note": "Use fetch(id) for full citeable content from a selected result."
}

Machine-readable resources

MCP manifest for the public endpoint, transport, authentication mode, and tool names.
MCP server card for public discovery metadata and read-only tool annotations.
OpenAPI 3.1 contract for the authenticated REST paths, request schemas, response schemas, and errors.

There is no static public schema file for the authenticated MCP tool catalog. After authentication, MCP clients discover the current tool schemas from the server. The public manifest and server card describe only the unauthenticated Public Knowledge MCP.

Get connected

Pick your AI assistant. The install steps and code panel update to match. Hosted clients use the connector at https://api.trakkr.ai/mcp; developer tools use the local setup shown in the code panel.

1
Get your connect token
Generate an MCP connect token from Settings → Developer. Tokens start with mcp_connect_ and are shown once.
MCP access is included on every paid plan. The direct REST API and sk_live_ keys remain Scale-only.
2
Create the Claude connector
Copy https://api.trakkr.ai/mcp into Claude's connector setup. When Claude opens the Trakkr authorization page, paste your MCP connect token there.
Claude MCP docs
3
Restart and ask away
Finish the Claude connection flow, then open a new chat. Try one of the prompts below to verify everything works.

Keep your connect token out of source control. Don't commit config files containing your token - add them to .gitignore or use your client's secret-input pattern.

Trakkr stores MCP activity logs for security, support, and product improvement: tool names, timing, status, short intent hints supplied by the assistant, and bounded redacted input and result previews. Full assistant conversations are not stored.

What you can ask

Your assistant picks the right tools automatically. Some prompts to try:

“How is my brand doing in AI search?”

“Which competitors are gaining ground?”

“What content should I create next?”

“Show me citation trends over the last quarter”

“Run a diagnosis on "best CRM for startups"”

“How does Perplexity describe my brand vs competitors?”

“Summarize my latest research run - visibility, top competitors, and the prompts I missed”

“Run a topic snapshot on "enterprise pricing pages" and tell me when it's done”

“What Reddit threads should I jump into this week?”

“List my active workflows and what fired in the last seven days”

“Show me unread notifications and mark them all as read”

“Wire a webhook to https://example.com/hook for visibility_changed and citation_gained”

“Compare brands ABC and XYZ on visibility, citations, and actions for the last 30 days”

“What's in my knowledge base and what's the most recent published article?”

“Across every brand I manage, what are the top 10 highest-impact actions right now?”

“How has my brand narrative drifted over the last quarter?”

“Using Trakkr research, what page types actually get cited in AI search?”

“Does llms.txt actually help me get cited?”

For multi-step queries, your assistant chains tools together - e.g. fetching your brand ID, then pulling scores, then comparing against a competitor - all in one turn.

Want to turn a finding into a shipped fix? The MCP Cookbook shows how to chain Trakkr with your other MCP servers (GitHub, your CMS, Slack, Search Console), so the gap Trakkr finds gets fixed in the same chat.

Available tools

76 tools across 12 groups. Each maps to an endpoint in the Trakkr API. Open a group to see parameters and views.

Resources

Resources pull Trakkr context into a chat with no tool call. In clients that support them, @-mention a resource (for example @trakkr://brand/<id>/briefing). Each is read-only and bounded, so reading one never triggers analysis or compute.

Your brand (paste-ready briefings)

Dense markdown you can @-attach to any chat, so the model gets your brand right.brand-book is the one to paste.

trakkr://brandThe index of your brands and their IDs, each linked to the briefings below. Start here to find a brand ID.

trakkr://brand/{id}/brand-bookPaste-ready context so any AI describes the brand right: what it is, how AI frames it now, what to set straight, and the proof.

trakkr://brand/{id}/snapshotLatest AI visibility: headline scores, a by-model table, and the prompts where you win and lose.

trakkr://brand/{id}/citation-gapsPrompts where AI cites a rival but not you, ranked, with who gets cited and what to publish.

trakkr://brand/{id}/promptsThe AI-search questions Trakkr is tracking for this brand, active and paused.

Your brand (live state, JSON)

trakkr://brandsThe brands you can access, with the IDs the per-brand resources need.

trakkr://brand/{id}/briefingHeadline visibility and trend, plus the open-action snapshot.

trakkr://brand/{id}/actionsThe brand top open actions, most impactful first.

trakkr://brand/{id}/changesWhat moved in the last week (the get_changes digest).

trakkr://brand/{id}/latest-reportThe most recent generated report.

Data & research

Every rival shows you your own numbers. These expose Trakkr’s public research on what actually gets brands cited, drawn live from our /data studies.

trakkr://dataIndex of the research briefings below.

trakkr://data/what-gets-citedWhich page types earn AI citations, and owned vs third-party.

trakkr://data/crawler-personalitiesWhat each AI bot actually fetches and reads.

trakkr://data/llms-txt-truthThe honest null result: llms.txt shows no citation lift.

trakkr://data/schema-advantageHow structured data correlates with getting cited.

trakkr://data/citation-decayHow fast AI citations fade.

trakkr://data/model-divergenceWhere AI models disagree on who to recommend.

trakkr://data/playbookThe synthesis: the rules for getting cited, with links.

Workflows

Workflows are server prompts that appear as slash commands in your assistant’s prompt menu the moment Trakkr connects. Each one chains the right tools and research into a single finished briefing, most important first. Arguments are optional; leave the brand out and the workflow resolves it for you.

/weekly-reviewbrand, period

Your Monday brief: visibility and trend, who is gaining, the biggest citation gaps, and the easy wins, as five tight bullets.

/competitor-teardownbrand, competitor

Where a rival beats you, where you beat them, and three concrete moves to close the gap.

/citation-gap-planbrand

The prompts you are losing, cross-referenced with the research on what actually gets cited, ending in a ranked plan of what to publish.

/content-briefbrand, topic

A citation-optimised outline plus the on-page schema to include for one piece, ready to hand to a writing tool.

/setup-trackingdomain

Onboarding in one chat: confirm the brand, set its market, then propose a starter set of prompts to track and create them on your say-so.

/trakkr-watchbrand_id, since

A watch playbook: pulls what moved for a brand, summarizes it plainly, and offers next steps. Pass the last cursor to see only what is new.

Troubleshooting

Package

trakkr-mcp

Version

0.17.0

Python

≥ 3.10

License

MIT

Code example

Claudehttps://api.trakkr.ai/mcp

Add Trakkr as a custom connector in Claude, then paste your MCP connect token when prompted.

1In Claude, open Customize > Connectors.
2Choose Add custom connector and use https://api.trakkr.ai/mcp as the remote MCP server URL.
3Click Connect, then paste your MCP connect token on the Trakkr authorization page when Claude prompts you.

Claude connector docs